Re: [Mesa3d-users] how to switch between hardware rendering and Mesa rendering

[Ethan G. <ee...@fa...>]

On Tue, 27 Oct 2009 11:33:21 +0800
Chia-I Wu <ol...@gm...> wrote:

> On Tue, Oct 27, 2009 at 01:45:56AM +0100, Florent Monnier wrote:
> > In my Linux system I have installed the proprietary drivers for my video card, 
> > But I would like to test a program with the Mesa rendering.
> > (Because the driver doesn't seems to support glsl and I would like to test a 
> > program with it)
> > So I have tryed to do what I have done before with other libs, which is I have 
> > installed Mesa with ./configure --prefix=/tmp/Mesa and then after the install 
> > I do: export LD_LIBRARY_PATH=/tmp/Mesa/lib
> > Then I run my program in the same session. With other libs this works fine, 
> > but here my program still run with the hardware accelerattion.
> > So you guess the question, how can I test a program with Mesa without 
> > uninstalling my hardware drivers ?
> With LD_LIBRARY_PATH set, you can
> 
> $ ldd <your-program>
> 
> to verify the libraries from mesa are used.  If it already is, and you
> are under X, you can run
> 
> $ glxinfo
> 
> to see if it is doing direct rendering, and which renderer is used.
> 
> If glxinfo reports direct rendering, you can make sure the software DRI
> driver is used by setting LIBGL_ALWAYS_SOFTWARE=1.
> 
> Another way is to use a Xlib-based GLX emulation.  You can configure
> mesa with
> 
> $ ./configure --with-driver=xlib
> 
> to have a libGL.so that does not talk GLX.

When I wanted to test software rendering I didn't know of the
LIBGL_ALWAYS_SOFTWARE varibale.  I configured Mesa like that, but
rather than mess with my system Mesa I configured it to install under
/opt like so:

$ ./configure --prefix=/opt/mesa-standalone --with-driver=xlib

Then when I want to test if something works with Mesa I use a Bourne
shell feature to set $LD_LIBRARY_PATH just for the one program:

$ LD_LIBRARY_PATH=/opt/mesa-standalone/lib program args


-- 
Ethan Grammatikidis

Those who are slower at parsing information must
necessarily be faster at problem-solving.